Perhaps no single concept is more pervasive and important in marketing than the notion of cause and effect. Marketing practitioners depend on it in the planning and implementation of programs designed to obtain responses from consumers. Ideas of advertising affecting sales, opinion leaders influencing the adoption of new products, or promotion activities producing interest and preferences for one’s wares all rely implicitly on mechanisms of cause and effect.1
Many of the logical fallacies in this book, such as Alleged Certainty, the Hasty Generalization, and Affirming the Consequent, are influenced by the cardinal sin of equating correlation with causation. How do you tell the difference between mere correlation and true causation?
Let’s first define “causation”: when one event will or can bring about another. One of the most common cause and effect relationships in marketing is reflected in the law of supply and demand—price reductions generally cause more consumers to purchase a product. There are three necessary criteria for saying that one variable is a cause of another: (1) temporal sequence, (2) concomitant variation, and (3) absence of other possible causes. All three conditions must be met to substantiate a causative relationship.
Temporal Sequence
If P causes Q, then P must occur before or at the same time as Q. Sometimes it’s difficult to say which comes first, as described by Malhotra:
For example, customers who shop frequently in a department store are more likely to have the credit card for that store. Also, customers who have a department store’s credit card are likely to shop there frequently. The time order of these variables—credit card ownership and frequent shopping—is not obvious. Did shopping precede credit card ownership? An understanding of the underlying phenomena associated with department store shopping might be necessary to accurately identify time order.2
Concomitant Variation
If P is a cause of Q, then the values of P and Q are correlated in the same direction. If the value of P increases, the value of Q increases, and vice versa. For example, if you expect that sales force quality is a cause of sales, then as the quality of the sales force increases (decreases), sales will increase (decrease) too, all other factors held constant.
Absence of Other Causal Factors
In a simple example of this condition, if P is hypothesized to cause Q, there are no other true causes of Q that happen to be correlated with both P and Q. Consider Figure 3.1 below.
In this example, an increase in social media spending (P) is hypothesized to cause an increase in sales (Q), which is denoted by the solid arrow from P to Q—solid arrows denote a possible causative relationship.
During this time period, however, the company also deployed a sales training program (R).There is dashed arrow from R to P denoting that these two variables are merely correlated and that changes in R do not cause changes in P.
Figure 3.1 Causal pathways
In contrast, there is solid arrow pointing from R to Q denoting that an increase in the quality of the sales force via a training program (R) could cause an increase in sales (Q).
But from the information provided, it’s impossible to infer what causes Q to increase. Do only changes in P cause changes in Q, only changes in R cause changes in Q, or do changes in both P and R cause changes in Q?
Most situations in marketing are similar to the above example—multiple events occur in markets at approximately the same time, which makes inferring causation problematic.
This does not, however, give one license to start using logical fallacies with abandon! For example, someone might employ the Affirming the Consequent fallacy to make the case that an increase in social media spending causes sales to increase by saying: “If our social media campaign were effective, then sales should increase. Sales increased. Therefore our social media campaign was effective.” This is poor reasoning because factors other than the social media campaign could have caused sales to increase.
Correlation, Causation, and Logical Fallacies
Short of running experimental research in which you hold constant or otherwise take into account all other factors except the one you are testing to determine causality, how do you avoid making erroneous causal claims? There is no simple answer to this question. In the above example, marketing research may reveal—for example, by interviewing customers—whether the sales training program or the social media campaign more likely explains an increase or decrease in sales. If you can’t conduct a marketing research study, you will need to construct alternative hypotheses—there is always more than one explanation for a set of data—and select the one that best explains the market behavior you’ve observed. This is called Inference to the Best Explanation (IBE). Among the competing hypotheses, select the strongest and most cogent one.
Interaction Effects and INUS Conditions
Identifying or inferring causation can become more problematic if interaction effects occur. An interaction effect is a condition in which a factor’s influence on a marketing outcome is affected by the presence of another factor. In the aforesaid example, maybe sales would increase only if a change were made both in social media and sales force training—no one action by itself would be decisive. There would be an interaction effect between social media and sales training when both have to be present to affect sales.
It gets worse! As described by Terry Grapentine in his book, Applying Scientific Reasoning to the Field of Marketing: Make Better Decisions, a single factor or group of factors—called a condition—may be either necessary or insufficient to influence a consumer to consider a brand. There are always some necessary conditions that have to be met for a consumer to consider a product. Clearly, if consumers cannot find a product, they cannot buy it. Product availability (X) is a necessary condition for purchase (Y). X, however, may be a necessary condition, but it may be insufficient by itself to bring about a purchase. The product also must be competitively priced—making X both a necessary and “insufficient condition” to motivate purchase.
Most causes of consumer purchasing behavior can be attributed to INUS conditions—Insufficient Necessary Unnecessary Sufficient, as explained by Richard Bagozzi (italics added):
As an example, let us examine the claim sometimes made by marketers that brand image (measured by the brand name) affects the perception of quality. When marketers make this claim they are not saying that the brand image is a necessary cause or condition for the attribution of quality. One may judge a product as high or low in quality without knowing the brand. Similarly, marketers are not claiming that the brand image is sufficient for the perception of quality since one must at least attend to, be aware of, and evaluate the brand name before such an attribution can be made. Rather, the brand image may be regarded as an INUS condition in that it is an insufficient but necessary part of a condition that is itself unnecessary but sufficient for the result. Many of the causal relations investigated by marketers are of this sort.3
The lesson to be learned with respect to not confusing correlation with causation when developing marketing strategies and tactics is this: Be mindful of the complexities of causality when making marketing arguments. Use evidence to support your causal claims.